The Fourth Extracellular Loop of the α Subunit of Na,K-ATPase

Na,K-ATPase is the main active transport system that maintains the large gradients of Na+ and K+ across the plasma membrane of animal cells. The crystal structure of a K+-occluding conformation of this protein has been recently published, but the movements of its different domains allowing for the cation pumping mechanism are not yet known. The structure of many more conformations is known for the related calcium ATPase SERCA, but the reliability of homology modeling is poor for several domains with low sequence identity, in particular the extracellular loops. To better define the structure of the large fourth extracellular loop between the seventh and eighth transmembrane segments of the α subunit, we have studied the formation of a disulfide bond between pairs of cysteine residues introduced by site-directed mutagenesis in the second and the fourth extracellular loop. We found a specific pair of cysteine positions (Y308C and D884C) for which extracellular treatment with an oxidizing agent inhibited the Na,K pump function, which could be rapidly restored by a reducing agent. The formation of the disulfide bond occurred preferentially under the E2-P conformation of Na,K-ATPase, in the absence of extracellular cations. Using recently published crystal structure and a distance constraint reproducing the existence of disulfide bond, we performed an extensive conformational space search using simulated annealing and showed that the Tyr308 and Asp884 residues can be in close proximity, and simultaneously, the SYGQ motif of the fourth extracellular loop, known to interact with the extracellular domain of the β subunit, can be exposed to the exterior of the protein and can easily interact with the β subunit.

Na,K-ATPase is the main active transport system that maintains the large gradients of Na ؉ and K ؉ across the plasma membrane of animal cells. The crystal structure of a K ؉ -occluding conformation of this protein has been recently published, but the movements of its different domains allowing for the cation pumping mechanism are not yet known. The structure of many more conformations is known for the related calcium ATPase SERCA, but the reliability of homology modeling is poor for several domains with low sequence identity, in particular the extracellular loops. To better define the structure of the large fourth extracellular loop between the seventh and eighth transmembrane segments of the ␣ subunit, we have studied the formation of a disulfide bond between pairs of cysteine residues introduced by site-directed mutagenesis in the second and the fourth extracellular loop. We found a specific pair of cysteine positions (Y308C and D884C) for which extracellular treatment with an oxidizing agent inhibited the Na,K pump function, which could be rapidly restored by a reducing agent. The formation of the disulfide bond occurred preferentially under the E2-P conformation of Na,K-ATPase, in the absence of extracellular cations. Using recently published crystal structure and a distance constraint reproducing the existence of disulfide bond, we performed an extensive conformational space search using simulated annealing and showed that the Tyr 308 and Asp 884 residues can be in close proximity, and simultaneously, the SYGQ motif of the fourth extracellular loop, known to interact with the extracellular domain of the ␤ subunit, can be exposed to the exterior of the protein and can easily interact with the ␤ subunit.
Na,K-ATPase is a membrane protein present in all animal cells. It belongs to the PII-type ATPase family, a family that includes the gastric and nongastric H,K-ATPase and sarcoplasmic reticulum Ca 2ϩ -ATPase (SERCA) 3 and the plasma mem-brane Ca 2ϩ -ATPase (1). Na,K-ATPase uses the energy provided by ATP hydrolysis to drive three Na ϩ ions out of the cell and two K ϩ ions into the cell. This enzyme is responsible for maintaining the electrochemical gradient of Na ϩ and K ϩ ions across the plasma membrane, which is necessary for regulating cell volume, for maintaining the resting membrane potential in excitable cells, and for many secondary active transport systems. The functional enzyme consists of an ␣ subunit (ϳ110 kDa) with 10 transmembrane segments (M1-M10) and a ␤ subunit (ϳ55 kDa) with a single transmembrane segment and a large glycosylated C-terminal ectodomain (2). The ␣ subunit contains the ATPase catalytic site, the binding sites for cations and for ouabain, a specific inhibitor of Na,K-ATPase, whereas the ␤ subunit is required for stabilization of the ␣ subunit and transit from the endoplasmic reticulum to the cell membrane (3). The cation transport mechanism by Na,K-ATPase across the plasma membrane is based on a cyclic scheme involving two main conformations, in which cation-binding sites are accessible either from the intracellular side (in the E1.ATP conformation) or from the extracellular side (in the E2-P conformation).
The crystal structure of Na,K-ATPase has been recently published (Protein Data Bank code 3B8E) (4). Na,K-ATPase was crystallized in the presence of K ϩ and MgF 4 2Ϫ , yielding a conformation with two K ϩ ions occluded in their binding sites in the transmembrane part of the protein, corresponding to the state just preceding the final release of P i (4). To facilitate the comparison with the published structure, we have adopted the 3B8E.pdb residue numbering scheme. In previous work, we have shown that the second (L3-4) and third (L5-6) extracellular loops linking the transmembrane segments have important roles in the control of the extracellular gate of the occlusion mechanism (5,6). In the present work we are interested in the role of the fourth extracellular loop, linking the seventh and eighth transmembrane segments (L7-8) and its relationship with the other parts of the extracellular domain.
The structure of the sarcoplasmic reticulum Ca 2ϩ -ATPase has been determined at high resolution under several conformational states with various ligands (7)(8)(9)(10)(11). Because of the sequence homologies of SERCA and Na,K-ATPase (globally ϳ28% amino acid sequence identity and 46% similarity), it has been possible to obtain structural models of Na,K-ATPase, which have helped to provide a better understanding of cation transport by Na,K-ATPase (6,(12)(13)(14)(15). All models predicted that, similarly to SERCA, the ␣ subunit of Na,K-ATPase includes 10 transmembrane segments (M1-M10) and that the extracellular part of this subunit is formed by five loops connecting pairs of transmembrane segments (L1-2 connecting M1 and M2, L3-4 connecting M3 and M4, and so on), and the predictions of these models are largely confirmed by the recently published structure (4).
A good correlation is observed between the model and experimental results obtained by functional studies of site-directed mutants, for instance for cation-binding sites, implicating M4, M5, and M6 for the K ϩ and two of the three Na ϩ -binding sites, and M9 for the third Na ϩ -binding site (13,16,17). However, the loops that form the extracellular parts of SERCA and Na,K-ATPase have more divergent sequences, with significant difference in length in case of L1-2, L3-4, and L7-8 loops in particular. The homology modeling process yields much less reliable results for these domains of the protein. In addition, the ␣ subunit of Na,K-ATPase is always associated with the ␤ subunit, and there is solid experimental evidence for a direct interaction of the large extracellular domain of the ␤ subunit and the L7-8 loop of the ␣ subunit, more precisely with a specific motif (SYGQ) present in this loop (18,19).
The specific inhibitors of Na,K-ATPase, such as ouabain, are known to bind from the extracellular side of the membrane, and their binding site seems to include the extracellular loops L1-2, L3-4, L5-6, and L7-8 (20 -24).
In the crystal structure of the SERCA, stabilized in the E2 cation empty conformation by the presence of thapsigargin, the second loop connecting M3 and M4, L3-4 comes into close contact with the fourth loop linking M7 and M8 (L7-8) (8). Because L5-6 is much shorter and does not come out of the membrane, it seems that L3-4 and L7-8 can come into contact with each other, leaving a rather large vestibule that could accommodate a passage for cation as well as a pocket for ouabain binding in the E2-P conformation. But no functional evidence has been provided concerning this possible interaction between the second and the fourth extracellular loops in Na,K-ATPase ␣ subunit.
To improve our knowledge of the structure of the extracellular domain of Na,K-ATPase, we reasoned that the demonstration of a direct interaction of residues in L3-4 and L7-8 could provide a strong geometrical constraint on the structure of these loops and greatly enhance our understanding of conformational changes of the pump structure taking place during the E2 phase of the pump cycle. For this purpose, we used sitedirected mutagenesis and functional measurements. Double cysteine mutants were generated, with a first cysteine mutation introduced in L3-4 at a position (Tyr 308 ) predicted by our previous work (6) and confirmed by the new crystal structure (4) to be located close to L7-8 and a second cysteine mutation in the L7-8 loop (positions Trp 883 to Asp 893 ), in the segment preceding the ␤ subunit association site ( 894 SYGQ). We then compared the functional effects of oxidizing and reducing agents on the double cysteine mutants, single cysteine mutants and wild type Na,K-ATPase. Our results indicate that the pair of residues Tyr 308 and Asp 884 is close enough (and more so in the E2-P conformation) so that when both mutated into cysteine they can be linked by a disulfide bridge when exposed to an oxidizing agent. Based on the structural constraint provided by this observation, we performed conformational space search studies that confirm the possibility of creating a disulfide bridge between Y308C and D884C without altering the rest of the Na,K pump structure.

MATERIALS AND METHODS
Oocytes-Female Xenopus laevis were anesthetized with tricaine MS 222 (2 g/liter; Sandoz, Basel, Switzerland). Parts of the ovaries were removed through a 1-cm ventral incision. These procedures were approved by the "Service Vétérinaire Cantonal" of the State of Vaud. Fragments of ovaries were treated with collagenase, and oocytes were incubated at 19°C in a modified Barth's solution containing 85 mM NaCl, 2.4 mM NaHCO 3 , 1 mM KCl, 0.8 mM MgSO 4 , 0.3 mM CaNO 3 , 0.4 mM CaCl 2 and 10 mM HEPES (pH 7.4), and supplemented with 10 mg/ml penicillin and 5 mg/ml streptomycin.
Site-directed Mutagenesis-Cysteine mutants were generated using the ␣1 subunit of the rat Na,K-ATPase subcloned into the pSD5 vector. The mutants were generated using the QuikChange site-directed mutagenesis kit from Stratagene (La Jolla, CA). All of the mutations were confirmed by a sequencing reaction. Mutant rat ␣1 subunit as well as wild type rat ␣1 and ␤1 subunit cRNAs were synthesized by in vitro transcription as previously described (25).
Expression of Rat Na,K-ATPase in Xenopus Oocytes-Ten ng of ␣ subunit and 1 ng of ␤ subunit cRNA were mixed and coinjected in a total volume of 50 nl into stage V-VI X. laevis oocytes prepared as described earlier (26). After injection, the oocytes were incubated for 3 or 4 days in modified Barth's solution and loaded with sodium by a 2-h incubation in a K ϩ -free and Ca 2ϩ -free medium containing 80 mM Na ϩ and 0.5 mM EGTA. They were then incubated overnight in a K ϩ -free amphibian Ringer solution containing 0.2 M ouabain to inhibit the endogenous Xenopus Na,K-ATPase. This concentration of ouabain allows selective study of the activity of the exogenously expressed rat Na,K pump, which is moderately ouabain-resistant (27).
Electrophysiological Measurements-The oocytes were loaded with Na ϩ and were studied using the two-electrode voltage-clamp technique. Voltage and current were recorded with a TEV-200 voltage clamp (Dagan, Minnneapolis, MN) and analyzed with the pCLAMP data acquisition package (Axon Instrument, Union City, CA). All of the experiments were performed at room temperature (22-25°C).
In each oocyte the Na,K pump current was activated by addition of 10 mM external K ϩ to a previously K ϩ -free solution (see composition of the control solutions below), and the ouabainsensitive current was measured by addition of 2 mM ouabain in the 10 mM K ϩ solution. The currents were measured at a holding membrane potential of Ϫ50 mV. The Na,K pump current was calculated as the difference between the steady-state currents in the presence and in the absence of extracellular K ϩ and the current inhibited by ouabain as the difference between steady-state currents recorded in the presence of extracellular K ϩ before and after inhibition by 2 mM ouabain.
Ouabain Affinity Measurements-The apparent affinity of ouabain was measured by recording the inhibition of the 10 mM K ϩ activated current by 0.1, 0.3, 1.0, and 3.0 mM ouabain. Oua-bain inhibition constant (K I ) was obtained for each oocyte as the best fitting parameter to the ouabain concentration-inhibition curve.
Effects of Copper-Phenanthroline and Dithiothreitol on Na,K-ATPase Function-To study the functional effect of the presence of a disulfide bridge in the extracellular domain of Na,K-ATPase, we used the oxidizing reagent, copper-phenanthroline at a concentration of 0.3 mM Cu 2ϩ and 0.9 mM phenanthroline and the reducing reagent dithiothreitol (DTT) at a concentration of 2.5 mM. The function of Na,K pump before and after exposure to the oxidizing or the reducing agent was measured as the K ϩ -induced current, calculated by subtracting the current measured in the K ϩ -free solution from the current measured in the presence of K ϩ .
To test for the existence of a spontaneously formed disulfide bridge, the induced current was first measured before and after a 10-min exposure to 2.5 mM DTT. Then the K ϩ -induced current was measured before and after a 10-min exposure to copper-phenanthroline. Finally, to control for the specificity of the copper-phenanthroline effects, DTT was applied again for 10 min, and the K ϩ activated current was measured again.
To investigate the possibility that the distance between the two cysteine residues could change according to the conformational state of the Na,K pump, copper-phenanthroline was added either in a 100 mM Na ϩ solution containing 10 mM of external K ϩ , a condition that favors the E1 conformation of the Na,K-ATPase, or in a Na ϩ -free and K ϩ -free solution, a condition that favors the E2-P conformation (6).
Reagents and Solutions-The composition of the K ϩ -free Na ϩ -containing control solution was 92.4 mM Na ϩ , 0.82 mM Mg 2ϩ , 5 mM Ba 2ϩ , 0.41 mM Ca 2ϩ , 10 mM tetraethyl ammonium, 22.4 mM Cl Ϫ , 2.4 mM HCO 3 Ϫ , 10 mM HEPES, 80 mM gluconate, pH 7.4, K ϩ channel blockers (Ba 2ϩ , tetraethyl ammonium) were used to minimize the effect of K ϩ on K ϩ channel currents. For the K ϩ -free and Na ϩ -free solution, Na ϩ gluconate was replaced by N-methyl-D-glucamine (NMDG) chloride resulting in the following solution: 92.4 mM NMDG, 0.82 mM Mg 2ϩ , 5 mM Ba 2ϩ , 0.41 mM Ca 2ϩ , 10 mM tetraethyl ammonium, 102.4 mM Cl Ϫ , 10 mM HEPES, pH 7.4. For the potassium containing solutions used to activate the Na,K pump, K ϩ gluconate (from a 1 M stock solution) was added to achieve a 10 mM K ϩ concentration in the case of the Na ϩ -containing control solution and 5 mM K ϩ in the case of the Na ϩ -free solution.
Ouabain (Fluka, Buchs, Switzerland) was added in the K ϩ -free amphibian Ringer solution at a concentration of 0.2 M from a 2 mM stock solution in Me 2 SO to inhibit the endogenous Xenopus Na,K-ATPase. For the 2 mM concentration experimental solutions, ouabain was directly dissolved in the final solution.
Copper-Phenanthroline-Phenanthroline (Sigma) was added at a concentration of 0.9 mM from a 900 mM stock solution in methanol conserved at Ϫ20°C. Copper sulfate (Sigma) was added at a concentration of 0.3 mM from a 300 mM stock solution. DTT (threo-1,4-dimercapto-2,3-butanediol) (Sigma) was added at a concentration of 2.5 mM from a 1 M stock solution kept on ice.
Data Presentation and Statistics-Because the expression of exogenous Na,K-ATPase in oocytes is variable, current ampli-tudes are reported as relative to the outward current induced by 10 mM K ϩ measured at Ϫ50 mV at the beginning of the experiment. We have shown earlier that this measure yields an accurate estimation of the number of active Na,K pumps at the surface of the oocyte (28). The results are reported as the means Ϯ S.E. (n ϭ number of measurements). The Student's t test for unpaired data were used for statistical comparison between wild type and mutant groups, and the Student's t test for paired data are used for comparison of different responses obtained on the same oocyte (before and after oxidizing reagent treatment, for instance). In case of comparisons between several groups, a one-way analysis of variance followed by a Bonferoni post-test was used. Mean differences with a value of p Ͻ 0.05 were considered statistically significant.
Molecular Modeling Procedures-The recently published crystal structure of Na,K-ATPase (Protein Data Bank code 3B8E) was used to investigate the possibility of creating disulfide bridge between residues of L3-4 and L7-8 loops of the pump. The loops conformation was modeled ab initio using simulated annealing protocol of the MODELLER program version 6.2 (29) The method was used to search the conformational space of the L3-4 and L7-8 loops, with the rest of the structure kept fixed. The residue Leu 302 was considered the first residue of L3-4 loop, and Thr 309 was considered the last one. In the case of L7-8 loop the Phe 871 was considered the first loop residue, and Gln 897 was considered the last one. An 4.5 Å harmonic constraint with standard deviation value 0.1 Å was set on the distance between the C␣ atoms of Tyr 308 and Asp 884 residues. This constraint was used to restrict the conformational space and search loop structures compatible with a disulfide bridge formation. 3000 conformations were generated and filtered using the MODELLER objective function; the 700 structures of the lowest MODELLER pseudo-energy value were selected for further study.
The resulting models were clustered on the basis of mutual similarity using the cluster.pl program from MMTSB tool set (30). The Atomic Non-Local Environment Assessment (ANOLEA) potential was used to assess the conformations of the loops. The ANOLEA energy profiles and the average ANOLEA scores for each conformation were calculated (31). The final conformation was chosen in the most populated cluster as that with the best ANOLEA score.

RESULTS
Functional Expression of Cysteine Mutants-The Na,K pump function measured as the outward current induced by 10 mM K ϩ and the ouabain-sensitive current (2 mM ouabain in the 10 mM K ϩ solution) were measured and compared in the wild type Na,K-ATPase, the single cysteine mutant Y308C and doublecysteine mutants in which any one of the residues from Trp 883 to Asp 893 was individually mutated to a cysteine in addition to the Y308C mutation (Fig. 1A). Except for the double cysteine mutant Y308C/W883C, for which no functional expression could be detected, all other double cysteine mutants showed a significant K ϩ -induced current, ranging from 43.4 Ϯ 3.2 nA, n ϭ 16 (Y308C/R886C) to 163.4 Ϯ 17.7 nA, n ϭ 10 (Y308C/ V891C), values to be compared with 154.6 ϩ 18.1 nA, n ϭ 22 for the wild type Na,K pump.
To test the hypothesis that the lack of functional activity of the Y308C/W883C mutant was due to the spontaneous formation of a disulfide bridge, we measured the K ϩ -activated current of this mutant before and after a 10-min exposure to 2.5 mM DTT. No K ϩ activated outward current could be detected before (Ϫ3.7 ϩ 0.5 nA, n ϭ 12), and DTT exposure had no effect (Ϫ3.9 ϩ 0.7 nA, n ϭ 12) (the small inward current is due to K ϩ flowing through incompletely blocked K ϩ channels).
The effects of 2 mM ouabain were also tested in these mutants. For the Y308C single mutant, the ouabain-sensitive current was significantly smaller than the K ϩ -induced current. The ratio of the ouabain-sensitive current to the K ϩ -induced current (I ouab /I K ; Fig. 1B) was 0.92 Ϯ 0.02, n ϭ 22 for the WT Na,K-ATPase and 0.67 Ϯ 0.03, n ϭ 16 for the Y308C mutant. The I ouab /I K ratio was similar to the Y308C mutant for all the double cysteine mutants except for two of them, the Y308C/ D884C and Y308C/R886C mutants. The I ouab /I K ratio of the Y308C/R886C mutant (1.04 Ϯ 0.06, n ϭ 16) was close to 1.0, similar to that of the wild type with, and significantly larger (p Ͻ 0.001) than the Y308C single mutant. The Y308C/D884C mutant had the particularity to have the smallest ouabain-sensitive current with a ratio of 0.37 Ϯ 0.04, n ϭ 10, a value significantly lower (p Ͻ 0.01) than that of the wild type or that of the Y308C single mutant, which means that the double cysteine mutant Y308C/D884C was more resistant to ouabain than the wild type Na,K-ATPase and the single cysteine Y308C mutant. This is probably due to a ouabain resistance induced by a change in the ouabainbinding site structure. It has been shown earlier that some residues of the loop linking M7 and M8, Arg 880 (32) and Asn 889 (33), are involved in ouabain binding. To confirm that the difference in K ϩ -induced versus ouabain-sensitive current were due to differences of ouabain affinities, the inhibitory constant (K I ) of ouabain was measured in the presence of 10 mM K ϩ in the wild type enzyme, the Y308C mutant, and the first four double cysteine mutants. The results, shown in Fig. 1C, indicate that all the Y308C mutants had a slightly lower affinity for ouabain than the wild type (120 ϩ 10 M, n ϭ 13) and that the Y308C/D884C mutant had a much reduced apparent affinity (5.1 ϩ 1.1 mM, n ϭ 10).
In summary we observed that, except for the Y308C/W883C double cysteine mutant, which was not studied further, the mutants were well expressed, and most of them displayed a small degree of resistance to ouabain, except for the Y308C/D884C mutant, which was clearly more ouabain-resistant, with 2 mM ouabain inhibiting less than half of the Na,K pump current.
Copper-Phenanthroline and DTT Effects on Double Cysteine Mutants in the E1/E2 Conformations of Na,K-ATPase-To detect the presence of a spontaneously formed and functionally important disulfide bond between cysteine residues in the extracellular domain of Na,K-ATPase ␣ subunit, the effect of DTT was first tested on all mutants. A 10-min exposure to DTT had no significant effect neither on the wild type Na,K-ATPase not or any of the single or double mutants tested (data not shown).
To detect oxidizing agent-induced formation of a disulfide bond between cysteine residues introduced in the ␣1 subunit of Na,K-ATPase, all of the mutants were tested by recording the effect of copper-phenanthroline on the activation by extracellular K ϩ . Copper-phenanthroline was first perfused for 10 min at concentrations of 0.3 and 0.9 mM and was tested under conditions favoring the E1 or the E2-P conformations of Na,K-ATPase by using a solution containing 100 mM of external Na ϩ and 10 mM of external K ϩ (favoring E1) or a Na ϩ and K ϩ -free (favoring E2-P) extracellular solution. The specificity of the copper-phenanthroline effect was controlled by perfusing a DTT solution on the same oocyte during 10 min at a concentration of 2.5 mM either in a solution promoting the E1 confor- were measured for oocytes injected with cRNA of rat ␤1 and ␣1 subunits of the WT, the Y308C mutant, and the double cysteine mutants of the ␣1 subunit of Na,K-ATPase from Y308C/W883C to Y308C/D893C. The Y308C/ W883C mutant had no detectable K ϩ -activated or ouabain-sensitive activity. The currents were recorded at Ϫ50 mV. Positive current values indicate an outward current. There were between 9 and 22 measurements in each group. B, the ratio of the ouabain-sensitive (2 mM) current and the K ϩ -activated current was calculated for each oocyte. Several mutants had a slightly lower ratio than the wild type, but only the Y308C/D884C mutant had a significantly lower ratio when compared with the Y308C mutant (p Ͻ 0.01, indicated by an asterisk). The error bars represent the S.E.

JOURNAL OF BIOLOGICAL CHEMISTRY 27853
mation or in a solution promoting the E2-P conformation of Na,K-ATPase. Fig. 2 (A and B) shows the amplitude of the K ϩ -activated current after copper-phenanthroline exposure (black bars) and after subsequent DTT exposure (white bars), normalized to the current values measured before copper-phenanthroline exposure, for the wild type, the single cysteine mutant Y308C, and all the double cysteine mutants from Y308C/D884C to D893C.
As shown in Fig. 2A, for experimental conditions favoring the E1 conformation, after a 10-min copper-phenanthroline exposure (black bars), the K ϩ -activated current was decreased between ϳ40 and 60% for all the double cysteine mutants, whereas the K ϩ -activated current of the wild type Na,K pump was not significantly inhibited.
Except for the Y308C/D884C mutant, the K ϩ -activated current was not significantly modified by subsequent exposure to DTT (black bars), suggesting that the decrease observed after copper-phenanthroline with all mutants (except Y308C/D884C) was probably not due to disulfide bond formation by copper-phenanthroline because it could not be reversed by exposure to the reducing agent DTT. In contrast, for the Y308C/D884C mutant, the K ϩ -induced current of this mutant was first reduced to 0.48 Ϯ 0.03, n ϭ 11 after copper-phenanthroline, (p Ͻ 0.001, when compared with before copper-phenanthroline or compared with wild type) but then was reactivated to 0.82 Ϯ 0.05, n ϭ 11, of its initial value after DTT (p Ͻ 0.001 when compared with the value recorded after copper-phenanthroline exposure or to the values in wild type or Y308C single mutant).
When the copper-phenanthroline exposure was performed in conditions favoring the E2-P conformation (Fig. 2B), the results were qualitatively similar. Except for the Y308C/D884C double mutant, the functional activity for the single and all the double cysteine mutants was The bar graphs report the amplitude of the K ϩ -activated current after copper-phenanthroline exposure and after subsequent DTT treatment. The K ϩ -activated current measured at Ϫ50 mV after exposure to 0.3/0.9 mM copper-phenanthroline (C-P) for 10 min is reported as the black bars, and the reversibility of the copper-phenanthroline effect, tested by measuring K ϩ -activated current after exposure to 2.5 mM DTT for 10 min, is shown as the white bars. The current values are normalized to the current measured before exposure to copper-phenanthroline. A, copperphenanthroline and DTT were perfused with 100 mM of external Na ϩ and 10 mM of external K ϩ solution to shift the equilibrium toward the E1 conformation of Na,K-ATPase. Exposure to copper-phenanthroline induced a decrease of the Na,K pump current in all cysteine mutants, when compared with the wild type protein, but none of the double cysteine mutants showed a DTT-induced recovery, except for the Y308C/ D884C mutant for which their was a significant (p Ͻ 0.001, indicated by an asterisk) increase upon DTT treatment. B, copper-phenanthroline and DTT were perfused in a Na ϩ -and K ϩ -free solution to shift the equilibrium toward the E2-P conformation of Na,K-ATPase. Exposure to copper-phenanthroline induced a large decrease of the Na,K pump current in the Y308C/D884C mutant (p Ͻ 0.001, indicated by an asterisk). None of the double cysteine mutants showed a DTT-induced recovery, except for the Y308C/D884C mutant for which their was a significant (p Ͻ 0.001, indicated by two asterisks) increase upon DTT treatment. C and D, the K ϩ -activated current measured at Ϫ50 mV after exposure to 0.3/0.9 mM copperphenanthroline for 10 min was reported (white bars), and the reversibility of the copper-phenanthroline effect was tested by measuring K ϩ -activated currents after exposure to 2.5 mM for 10 min (black bars). The bar graphs report the amplitude of the K ϩ -activated current after copper-phenanthroline exposure and after subsequent DTT treatment. The current values are normalized to the current measured before exposure to copper-phenanthroline. Copper-phenanthroline and DTT were perfused either with 100 mM of external Na ϩ and 10 mM of external K ϩ solution (C) to shift the equilibrium toward the E1 conformation of Na,K-ATPase or in a Na ϩ -and K ϩ -free solution (D) to promote the E2-P conformation of Na,K-ATPase. Between 5 and 13 measurements were performed for each condition. The error bars represent the S.E. In the E1 conformation (C), the recovery upon DTT treatment was significantly larger in the Y308C/D884C mutant than in the other mutants or with the control experiment (*, p Ͻ 0.01, paired student t test). In the E2-P conformation (D), the effect of copper-phenanthroline was statistically significant (*, p Ͻ 0.001, student t test) for the double cysteine Y308C/D884C mutant compared with the Y308C/D884C control perfused with the solvents of copper and phenanthroline, water, and methanol, respectively, and the recovery upon DTT treatment was also highly significant (**, p Ͻ 0.001, paired Student's t test).
reduced to ϳ60 and 70% (black bars) compared with ϳ90% in wild type, and the 10-min exposure to DTT either had no effect or induced an additional reduction of the Na,K pump functional activity for all the mutants (white bars). In contrast the Y308C/D884C mutant was very strongly inhibited by copperphenanthroline, to 12 Ϯ 3%, n ϭ 10, of the initial control value (p Ͻ 0.001 for comparison with wild type, Y308C single mutant and all the other double mutants). Subsequent DTT exposure induced a large recovery to a level of Na,K pump activity of 63 Ϯ 5%, n ϭ 10 (p Ͻ 0.001 when compared with the value after copper-phenanthroline). This can be considered as a full recovery because the current recorded at this point was similar to that observed in the wild type. This reduction correspond to a usually observed rundown of the Na,K pump activity in an experiment lasting ϳ30 min. For the Y308C/D884C mutant, the decrease of Na,K pump activity induced by copper-phenanthroline was significantly more pronounced in conditions favoring the E2-P conformation (activity reduced to 12 Ϯ 3%, n ϭ 10) than in conditions favoring the E1 conformation (activity reduced to 48 Ϯ 3%, n ϭ 11, p Ͻ 0.001), whereas the reverse was true for all of the other double mutants. In addition, of all double cysteine mutants tested with the copper-phenanthroline, the Y308C/D884C was the only one to present a decrease in its functional activity that nearly fully recovered upon DTT treatment.

Copper-Phenanthroline and DTT Effect on the Y308C/ D884C Mutant and the Corresponding Single Cysteine
Mutants-In a second set of experiments, the effects of copper-phenanthroline and DTT were compared between the Y308C/D884C double cysteine mutant, the corresponding single mutants, the Y308C and the D884C, and the wild type. In addition, a time control was performed by measuring the effects of a solution containing the phenanthroline solvent (methanol) and copper solvent (H 2 O) on the Y308C/D884C mutant. Fig. 2 (C and D) shows the amplitude of the K ϩ -activated current after copper-phenanthroline exposure (white bars) and after subsequent DTT exposure (black bars), normalized with the current values measured before copper-phenanthroline exposure, for the wild type, the double cysteine Y308C/D884C mutant, the single mutants, namely the Y308C and the D884C, and for the Y308C/D884C mutant in the control solution. Fig. 2C shows the results of the copper-phenanthroline and DTT effects obtained in the conditions favoring the E1 conformation of the Na,K-ATPase (exposure in the presence of Na ϩ and K ϩ ). The copper-phenanthroline exposure resulted in a decrease of the K ϩ -activated current ϳ0.39 Ϯ 0.04, n ϭ 10 for the Y308C, 0.09 Ϯ 0.04, n ϭ 11 for the D884C, 0.52 Ϯ 0.03, n ϭ 11 for the double cysteine mutant Y308C/D884C, 0.17 Ϯ 0.06, n ϭ 5 for the Y308C/D884C control, whereas no decrease was observed for the WT. After copper-phenanthroline treatment, DTT was tested on the same oocytes, and the Y308C/D884C displayed a significant increase of the K ϩ -activated current, whereas it was not significantly different of that measured after copper-phenanthroline exposure for the wild type or the Y308C and the D884C mutants or in the control experiments where the double cysteine mutant was not exposed to the reagents.
Copper-phenanthroline was next tested in conditions favoring the E2-P conformation of the Na,K-ATPase (copper-phenanthroline exposure in Na ϩ -free and K ϩ -free solution); the results are shown in Fig. 2D. A slight decrease of the K ϩ -induced current appeared after copper-phenanthroline exposure for the WT and the Y308C and D884C mutants to 96 Ϯ 3%, n ϭ 9, 71 Ϯ 4%, n ϭ 13 and 98 Ϯ 6%, n ϭ 9, respectively. The double mutant in which the Tyr 308 and the Asp 884 residues were replaced by cysteine displayed a strong reduction of its functional activity following copper-phenanthroline treatment because its K ϩ -activated current was reduced to 9 Ϯ 5%, n ϭ 10, a value significantly smaller than that observed with the same mutant (51 Ϯ 6, % n ϭ 5, p Ͻ 0.002) in the control experiment without copper sulfate and phenanthroline. DTT exposure fully restored the activity of the double Y308C/D884C mutant to a value similar to that observed in the control experiment, whereas it had no significant effect in the D884C mutant and produced a small additional decrease in the WT and the Y308C mutant.
Modeling-The ab initio modeling of the L3-4 and L7-8 Na,K-ATPase loops was performed. The E2 conformation of Na,K-ATPase pump was modeled, because according to our results the disulfide bridge was more readily formed under the conditions that favor this conformation.
The results of the conformational space search using the described distance restraint in the crystal structure in [2K ϩ ]E2⅐P state indicate that the disulfide bond between the Y308C and D884C residues is compatible with the rest of the structure, upon a possible conformational change of the loop; the distance constraint of 4.5 Å could, indeed, be satisfied in all of the sampled loops.
The selected loop conformation of Na,K-ATPase was chosen in the most populated cluster as the structure with the best ANOLEA score (see "Materials and Methods"). Its conformation with the disulfide bridge formed between the Y308C and D884C residues is shown in Fig. 3.

DISCUSSION
This study was carried out to bring additional structural information about the structure of the extracellular domain of the ␣ subunit of Na,K-ATPase during and the potential modifications of this structure during conformational changes.
Selection of Cysteine Mutations-The structure of SERCA under multiple confirmation indicate that the L3-4 and the L7-8 loops come in close contact in the E2-P conformation (8), and our previous modeling work using ab initio calculations for the short L3-4 loop (constituted by the residues 304 LILEYTW 310 ) (6) had led us to propose a precise structure for this loop, in which Tyr 308 , located in the middle of the loop, is brought close to the L7-8 loop in the E2 conformation (8.6 Å from backbone to backbone in E2, compared with 10.8 Å in E1). In addition the side chain of Tyr 308 is pointing toward the L7-8 loop.
The choice of the region to explore for a potential contact site in the L7-8 loop was more difficult. L7-8 is much longer than L3-4, and its length differs between SERCA (32 residues) and Na,K-ATPase (27 residues). This loop is known to contain a segment essential for the association between the ␣ and the ␤ subunit; a SYGQ motif in this segment is conserved in all Na,Kand H,K-ATPases ␣ subunit known to be strictly associated with a ␤ subunit, and it is not present in any known SERCA sequence (18,19). Indeed, Béguin et al. (34) provided evidence that association of the ␤ subunit with the SYGQ motif is necessary but not sufficient for the correct packing of the ␣ subunit, and a mutational analysis of a region encompassing Val 881 and Val 907 in the L7-8 loop revealed a significant influence of amino acid residues surrounding the 894 SYGQ and more particularly the 893 DSYGQQWTY 901 segment. Because of its close association with the ␤ subunit, we reasoned that these residues are not available for a possible interaction with the second extracellular loop, and thus we focused on the part of L7-8, from the Trp 883 to Asp 893 residue, preceding the ␤ subunitassociated domain. When the Na,K pump crystal structure became available, we could make a structure-based sequence alignment and see that this segment also comprises the residues of Na,K-ATPase (RWI 888 ) corresponding to residues of SERCA (HFM 874 ), which are the closest to the second extracellular loop in the cation-free E2-P SERCA Protein Data Bank structure 1IWO.
Exposure to DTT did not produce any consistent changes of the Na,K pump activity, neither in the wild type nor in any of the tested mutants. This observation indicates that there is no disulfide bridge present in the extracellular part of Na,K-ATPase that is critical for the function of the transporter. For the wild type protein, the crystal structure of the ␣ subunit indicates that the cysteine residues present in the extracellular part of the protein or close to it are not close to each other. Similarly from the same structure we could verify that the position chosen for the mutations are not closely located to any existing cysteine. Three disulfide bridges are known to be present in the extracellular domain of the ␤ subunit (35), but our results indicate that either they are resistant to DTT treatment or the reduction of these bridges has no obvious and immediate functional consequences.
A Functional Evidence of a Link between the Loops 3/4 and 7/8-Our main finding is that the Y308C/ D884C mutant was strongly inhibited by exposure to the oxidizing agent copper-phenanthroline and significantly more so when the exposure was performed under conditions where the Na,K-ATPase is preferentially maintained in the E2-P conformation than when the E1 conformation is favored. In addition this inhibition could be safely attributed to the formation of a disulfide bond because the inhibition was rapidly reversible upon treatment with the reducing agent DTT. The hypothesis that the effect of copper-phenanthroline was due to formation of a disulfide bond was also strongly supported by the observation that the two cysteine residues were required for the large inhibition and the reversibility by DTT. When no cysteine was present (wild type) or only one (Y308C mutant or D884C mutant), no inhibition or only a weak inhibition could be observed, and it was not reversible by DTT, suggesting a nonspecific effect of copper-phenanthroline. This effect was probably related to the presence of a cysteine residue at position 308, because it was not observed neither in WT nor in the single cysteine D884C mutant. ). Pairs of transmembrane segments are color-coded similarly in all four panels. The position of a disulfide bridge between the two side chains of substituting cysteine residues at position Tyr 308 and Asp 884 is indicated in gray. The side chains of the four residues (SYGQ) known to be essential for the interaction with the ␤ subunit are also shown. Graphics produced with the VMD application (37).
Concerning the Y308C/W883C mutant, the lack of effect of exposure to DTT does not support the hypothesis that the lack of functional activity is due to the spontaneous formation of a disulfide bridge. It has to be noticed that Trp 883 is the most highly conserved residue of the 10 that were mutated (conserved in all known Na,K-ATPase alpha subunits), and thus it is not surprising that this residue must have a very important role and that any mutation at this position would result in an altered function.
Although DTT exposure did not induce supplementary effect to what we have observed with copper-phenanthroline in mutants studied in the E1 conformation, a decrease was seen in the WT and the majority of mutants exposed in the E2-P conformation except for the Y308C/D884C, Y308C/R886C, Y308C/E892C, and the D884C mutants. The fact that this response was observed in most mutants suggests that this decrease of the functional activity is not related to the specific position of a cysteine residue and thus not to the action of DTT on a specific disulfide bond that includes the introduced cysteine but rather to a nonspecific effect or a general rundown of the Na,K-ATPase activity during the rather long duration of our experiments (30 -35 min).
The Link between the Loops 3/4 and 7/8 and the Ouabainbinding Site-Ouabain is a member of the cardiotonic steroids, which are mainly used in the treatment of congestive heart failure and arrhythmias. Ouabain is a specific inhibitor of the Na,K-ATPase and binds with a higher affinity the phosphorylated E2 form (36). The ouabain-binding pocket of the Na,K-ATPase has been reconstituted in the gastric H,K-ATPase by substituting seven amino acids, of which four are located at the top of M4 (20). Moreover, it has been found that glutamine residue Glu 889 located in the fourth extracellular loop is involved in the access and release of ouabain to and from its binding site (33). Therefore, the M4 segment and some residues of the loop 7/8 are important for the structure of the ouabainbinding site.
The double cysteine mutant Y308C/D884C presented a higher resistance to ouabain than the single cysteine mutants Y308C and the D884C. Indeed, the single cysteine mutants Y308C and D884C had the same degree of resistance to ouabain (0.67 Ϯ 0.03, n ϭ 16 for the Y308C, versus I ouab /I K ϩ 10 mM ϭ 0.68 Ϯ 0.04, n ϭ 14 for the D884C). The effects of these mutations on ouabain binding seemed to be additive because the double cysteine mutant Y308C/D884C was more resistant than each of the single mutants.
This observation suggests that Tyr 308 and Asp 884 can interact and are thus closely located. Taken together, these findings support that the proximity between the loops 3/4 and 7/8 in the E2-P conformation is important for the binding of ouabain.
In the recently published crystal structure of Na,K-ATPase, the residue in the fourth extracellular loop closest to Tyr 308 (in the second extracellular loop) is not Asp 884 , but rather Leu 879 . The distance from Tyr 308 to Asp 884 is 17.5 Å, which is too large for the formation of a disulfide bond. This apparent discrepancy can be easily understood considering that the transmembrane segments and the attached extracellular loops undergo rather large movements during the conformational changes of the pump cycle. The published crystal structure (4) was obtained after stabilization by the presence of MgF 4 2Ϫ instead of phosphate and in the presence of the K ϩ -congener Rb ϩ and corresponds to a confirmation with a closed external gate, occluding 2 K ϩ ions ([2K ϩ ]E2⅐P). Our experiments were designed to measure the reactivity to oxidizing agent under conditions where the majority of Na,K-ATPase is in the phosphorylated E2-P conformation without cations in the binding site and with an open extracellular gate. Comparison of our results with the published crystal structure thus suggests that the closing of the extracellular gate (the main molecular physiological change between the two conformations ([2K ϩ ]E2⅐P) and E2-P) is accompanied by a significant movement of the L7-8 extracellular loop.
At the same time we assume that changes in the pump transmembrane segments conformation between ([2K ϩ ]E2⅐P) state, represented in the crystal structure, and E2-P state, observed during our experiments, are less radical than changes in the conformation of the loop regions; thus we could use the Na,K-ATPase crystal structure in the molecular modeling studies.
The formation of a disulfide bridge between the two substituting cysteines at positions 308 and 884 indicate that in its native environment, these two positions are allowed to come in close contact, at a distance allowing for the formation of a S-S link, and the fact that the formation of this link occurs more readily when exposure to copper phenanthroline is performed in the absence of extracellular Na ϩ and K ϩ indicates that the 308 and 884 positions are more frequently in close proximity when the pump is in the E2-P conformation.
In conclusion, our results show that an oxidizing agent can induce the formation of a disulfide bond between two cysteines added in the second and the fourth extracellular loop, at the 308 and 884 positions, respectively. The formation of a disulfide bond was detected by a reduced function of Na,K pump upon treatment with copper-phenanthroline and recovery upon DTT treatment. These findings provide a functional evidence for a close proximity between the L3-4 and L7-8 loops that, when compared with the published crystal structure, indicate a significant movement of these loops during conformational changes.
The obtained model of L3-4 and L7-8 loops conformations provide additional validation of our experimental results by showing that Tyr 308 and Asp 884 residues can be in the close proximity while the pump is in the E2 state and that simultaneously the SYGQ motif of L7-8 loop can be exposed to the exterior of the protein, which makes possible its interaction with the ␤ subunit. This novel structural information could provide the basis for studies of the extracellular ion access site and ouabain-binding site.